Let $f:\mathbb{R}\rightarrow\omega_1$ be the "usual" surjection (send $r$ to the ordinal coded by $r$ according to some fixed reasonable coding system if such an ordinal exists, and to $0$ otherwise). Consider the following statement:
$(\%)\quad$ If $G\subseteq\omega_1^\omega$ is determined (as a game on $\omega_1$) then the "preimage" $\hat{G}=\{s\in\mathbb{R}^\omega: (f(s_i))_{i\in\omega}\in G\}$ is determined (as a game on $\mathbb{R}$).
This principle follows trivially from $\mathsf{ZF+AD}_\mathbb{R}$ or $\mathsf{ZFC}$. However, the $\mathsf{ZF+AD}$ situation is unclear to me: if we let $\Sigma$ be a winning strategy for (say) player $1$ in $G$, there is no obvious way to convert it to a winning strategy for $1$ in $\hat{G}$ since that would require a canonical choice of code for each ordinal.
Question: Does $\mathsf{ZF+AD}\vdash(\%)$?
I'm happy to look instead at mild strengthenings of $\mathsf{ZF+AD}$ (e.g. add $\mathsf{V=L(\mathbb{R})}$ or $\mathsf{AD}^+$).